A gene methylation marker combination and screening model for the screening of high-grade cervical lesions

Abstract

The present invention firstly discloses a gene methylation biomarker combination and a screening model for the screening of high-grade cervical lesions. By comparing differentially methylated positions (DMPs) in normal tissues and high-grade lesion tissues, and using the corresponding CpG islands as diagnostic biomarkers for research, the present invention screens out 12 optimal CpG islands with consistent characteristics through a random forest classifier, and establishes a screening model based on the CpG island level of cervical exfoliated cells. This model can perform risk assessment on patients with positive HPV and negative cytology, distinguish high-risk groups from low-risk groups, and recommend different follow-up or intervention measures clinically to achieve patient stratification and effectively improve the accuracy of cervical cancer screening; the screening model has the advantages of high sensitivity and specificity, and objective risk assessment results.

Description

Technical Field

[0001] The present invention belongs to the field of biotechnology, and particularly relates to a gene methylation biomarker combination and a screening device for screening high-grade cervical lesions. Background Art

[0002] Persistent infection with high-risk human papillomavirus (HPV) is the main cause of cervical cancer, and high-risk HPV can be detected in 99.7% of cervical cancers. High-risk HPV infection is also closely related to high-grade cervical squamous intraepithelial neoplasia (SIL), while low-risk HPV mainly induces condyloma acuminata in the external genitalia, skin and other parts, as well as low-grade cervical squamous intraepithelial neoplasia (SIL), etc. At present, the main method for cervical cancer screening clinically used is the combination of cervical liquid-based cytology examination and HPV virus detection. However, the cytological examination is greatly affected by humans, and the probability of missed diagnosis is relatively high. The limitations of HPV detection are as follows: (1) The detection result of this method cannot distinguish transient HPV infection and pathogenic persistent infection; (2) In women of childbearing age, especially women under 30 years old, the infection rate of HPV is high, and the clinical indication significance is weak. Relying solely on HPV screening, its low specificity is likely to cause waste of limited social and medical resources and over-treatment of patients.

[0003] Cervical tissue morphological diagnosis is the "gold standard" for cervical cancer diagnosis. However, morphological diagnosis highly depends on the personal ability of pathologists, has great subjectivity, and there is a large gap in the number of pathologists in China at present, far from meeting the clinical needs. Therefore, a means is needed to assist tissue morphological diagnosis and minimize the influence of subjective factors of pathologists on the diagnosis result. In recent years, with the research on epigenetics, DNA methylation, as one of the gene epigenetic modification methods, has gradually become an emerging diagnostic method for the detection of cervical precancerous lesions. Promoter methylation in the CpG island region may lead to the silencing of tumor suppressor genes, thereby affecting the tumor process. Compared with other biomarkers, DNA methylation has the advantages of high stability, low requirements for sample preservation, high compatibility with sample types, etc., and is very suitable as a diagnostic biomarker. A large number of methylation sites are also considered as biomarkers for cervical cancer screening.

[0004] In order to further improve the screening efficiency of cervical cancer, it is of great clinical significance to develop a set of DNA methylation biomarker combinations for screening high-grade lesions for cervical tissue or cervical cells. Summary of the Invention

[0005] To fill the gaps in the existing technology, the purpose of the present invention is to provide a gene methylation marker and a screening device for the screening of high-grade lesions of cervical tissue or cervical cells. By detecting the DNA methylation level of cervical exfoliated cells and predicting whether the specimen is a high-grade lesion according to the methylation level value of the CpG island, the purpose of auxiliary screening for cervical cancer is achieved.

[0006] The technical solution of the present invention to solve the above technical problems is as follows:

[0007] The first aspect of the present invention provides a gene methylation marker combination for the screening of high-grade cervical lesions. The methylation marker combination contains 12 CpG island regions, and their positions on the genome and the names of the genes / regions to which they belong are as follows:

[0008] (1) chr19:11959577-11960064, the region belongs to the intergenic region (IGR);

[0009] (2) chr18:70533965-70536871, the gene is NETO1;

[0010] (3) chr19:22805800-22806721, the region belongs to the intergenic region (IGR);

[0011] (4) chr20:21686199-21687689, the gene is PAX1;

[0012] (5) chr12:101603387-101603933, the gene is SLC5A8;

[0013] (6) chr11:132952538-132953307, the gene is OPCML;

[0014] (7) chr19:58609338-58609988, the gene is ZSCAN18;

[0015] (8) chr12:128751041-128753151, the gene is TMEM132C;

[0016] (9) chr8:110986113-110986983, the gene is KCNV1;

[0017] (10) chr9:139780406-139781527, the gene is TRAF2;

[0018] (11) chr8: 105478672 - 105479340, the gene it belongs to is DPYS;

[0019] (12) chr3: 5137548 - 5137977, the region it belongs to is the intergenic region (IGR);

[0020] Furthermore, the screening for high - grade cervical lesions uses cervical tissue or cervical exfoliated cells as samples.

[0021] The second aspect of the present invention provides a screening device prepared based on the above - mentioned gene methylation marker combination. The device includes a detection module and a prediction module;

[0022] Furthermore, the detection module includes genomic DNA extraction reagents, bisulfite, genomic DNA recovery reagents, PCR amplification reagents, and methylation chip hybridization reagents; the prediction module includes a constructed random forest classifier;

[0023] Furthermore, the construction method of the random forest classifier is as follows:

[0024] S1. Obtain normal cervical tissues, CIN1, CIN2, and CIN3 lesion tissues of the cervix, extract genomic DNA, perform bisulfite treatment and PCR amplification, hybridize the amplification products on a chip, scan the hybridization signals, obtain the signal values of each probe, calculate the β - value of each probe according to the formula, and normalize and correct the methylation β - value;

[0025] S2. Use the ChAMP software package (based on R language) to analyze the differentially methylated positions (DMPs) of the data, obtain the DMPs positions, and take the corresponding CpG islands and their numbers; use the 450K chip dataset GSE46306 as the test set and perform DMPs analysis in the same way; take the intersection of the differentially methylated CpG islands from the two datasets for subsequent analysis;

[0026] S3. Define the methylation level of a CpG island as the average β - value of the DMPs positions contained in the CpG island. Use the methylation level of the intersection CpG islands in step S2 as the independent variable and whether it is a high - grade lesion as the dependent variable for data modeling. Use the random forest (RF) model in the Scikit - learn software package (based on Python) to construct a classifier. After feature screening and model training on the above - mentioned intersection CpG islands, a classification model is obtained;

[0027] S4. Sort the correlation coefficients of each CpG island in the classification model in step S3, and further explore the number of CpG islands included in the model. Finally, an optimal combination containing 12 CpG islands is obtained;

[0028] S5. Construct a random forest classifier with 12 CpG islands and test its effect;

[0029] Specifically, the object of the present invention is to obtain the best classification effect with the least number of CpG islands;

[0030] Further, the calculation method of the β value in the step S1 is: signal intensity of methylation probe / (signal intensity of methylation probe + signal intensity of non-methylation probe + 100);

[0031] Further, when analyzing DMPs, the reference genome version is hg19;

[0032] Further, the detection is specifically to extract genomic DNA from cervical tissue or exfoliated cell samples to obtain the methylation β value of CpG islands;

[0033] Further, the prediction is specifically to input the methylation β values of each obtained CpG island into the constructed random forest classifier to obtain a prediction result;

[0034] Further, the prediction result takes the probability that the predicted tissue is a high-grade lesion as the output value, and the numerical range is 0 to 1; where 1 represents 100%, 0 represents 0%, and the larger the value, the more the tissue tends to be a high-grade lesion;

[0035] Further, when the prediction probability exceeds the evaluation threshold, the sample is classified into the high-risk group, and the patient himself / herself is a high-risk population, and intensive follow-up or direct clinical intervention is required;

[0036] Specifically, in the specific embodiment of the present invention, the evaluation threshold of the prediction probability is 0.51 (i.e., 51%), and at this threshold, the sensitivity and specificity of the model on the test set are 82.35% and 95.00% respectively.

[0037] The beneficial effects of the present invention are as follows:

[0038] The present invention for the first time provides an application and a screening model based on DNA methylation markers in cervical tissue or exfoliated cells in the screening of high-grade lesions. By comparing the differentially methylated sites (DMPs) in normal cervical tissue and high-grade lesion tissue, the differentially methylated sites with statistically significant differences are screened out. Through a random forest classifier, 12 best DMPs with consistent features are screened out, and a screening model at the methylation site level for screening cervical high-grade lesions is established. This model can further classify patients with positive HPV and negative cytology, predict cervical high-grade lesions, and effectively improve the accuracy of cervical cancer screening; the screening model has the advantages of high sensitivity and specificity, and an objective detection process. Description of the Drawings

[0039] Figure 1 Shown as the methylation β - value normalization graph of the main dataset GSE143752 and the test dataset GSE46306;

[0040] Figure 2 Shown as the learning curve of the RF model and the graph for determining the number of 12 best CpG islands;

[0041] Figure 3 Shown as the result graph of two - time parameter adjustment of the classification model;

[0042] Figure 4 Shown as the receiver operating characteristic curve of the screening model of the present invention in the test dataset;

[0043] Figure 5 Shown as the UCSC CpG island names, gene names and regions corresponding to 12 CpG islands. Detailed implementation manners

[0044] The following illustrates the present invention with examples, but does not limit the present invention. In the art, simple substitutions or improvements made by those skilled in the art to the present invention fall within the scope of the technical solutions protected by the present invention.

[0045] Example 1:

[0046] The method for detecting gene methylation expression levels from cervical tissues or cells provided by the present invention can well obtain DNA from samples and conveniently complete the detection.

[0047] A total of 104 cases in the normal vs cin1 group and 82 cases in the cin2 vs cin3 group of cervical exfoliated cell specimens were included in the present invention, and the total number of samples was 186 cases.

[0048] Genomic DNA was extracted from the samples. For the extracted DNA, the total amount and purity of the DNA were measured using an ultraviolet spectrophotometer to ensure that the DNA met the laboratory quality control requirements.

[0049] The extracted DNA was subjected to bisulfite treatment. A certain amount of DNA was incubated with the bisulfite solution at room temperature for 1 hour. After incubation, it was passed through a centrifugal column, and the treated DNA was purified and recovered.

[0050] 1 μg of DNA treated with bisulfite was taken for chip hybridization. The chip hybridization kit used the commercial kit HumanMethylation850 BeadChip kits from Illumina Company, USA, and the experimental steps were carried out according to the manufacturer's instructions.

[0051] After the chip hybridization is completed, the chip is placed in the chip scanner supporting by Illumina company of the United States to read the chip fluorescence signal data and obtain the idat file of the chip scan.

[0052] Import the obtained idat file into the ChAMP package (based on R language), and combine the grouping information to perform steps such as CpG island filtering, data normalization, data quality control, and principal component analysis. Then, according to the DMP analysis process of the ChAMP package, perform DMP analysis, and initially obtain 636 differential loci between the two groups. Further, intersect with the CpG islands in the 450K chip dataset GSE46306, and 286 CpG islands in the intersection part are included in the subsequent analysis. Import the above-optimized 286 CpG island loci, sample grouping information, sample β value matrix, etc. into the Scikit-learn package (based on Python). According to whether the expression trends of the 286 differential CpG islands are consistent in the two datasets, 72 CpG island differential loci are screened.

[0053] Use the Random Forest model to construct a classifier. Bring all 72 CpG island loci into the model for training, and it is found that when the number of included CpG islands is the first 12, the model can reach the optimal (score of 0.84). Including more CpG islands has little improvement on the overall model or even leads to a decline in the model effect, as Figure 2 shown. After obtaining and testing the influence of different model parameters on the model effect and determining the optimal parameters of the model, the classification model can be further optimized (score of 0.89), as Figure 3 shown.

[0054] Based on the above classification model parameters, draw a learning curve through the test set, and the cut-off value of the model can be obtained as 0.51. The prediction sensitivity and specificity can reach 82.35% and 95.00% respectively, and the area under its curve (AUC) is 0.88, as Figure 4 shown.

[0055] Information of the preferred 12 CpG island loci: the corresponding UCSC CpG island name, gene name and position attributes, as Figure 5 shown.

[0056] In actual use, only need to detect the β values corresponding to the preferred 12 CpG islands for any cervical tissue or exfoliated cells according to the process described in the present invention, and bring the β values of each CpG island into the RF classification model to obtain the prediction result of whether the sample is a high-grade lesion. The result is expressed as a probability, and the value range is 0-1, where 1 represents 100% and 0 represents 0%. The larger the value, the more the sample tends to be a high-grade lesion.

[0057] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the inventive concept of the present invention, several modifications and improvements can be made, and these all fall within the protection scope of the present invention.

Claims

1. Use of a reagent for detecting a combination of gene methylation markers for screening high-grade cervical lesions in the preparation of a product for diagnosing high-grade cervical lesions, characterized in that: The methylation marker combination includes 12 CpG island regions, the locations of which on the genome and the names of the genes / regions to which they belong are as follows: (1) chr19:11959577-11960064, the region belongs to the intergenic region (IGR); (2)chr18:70533965-70536871, the gene is NETO1; (3) chr19:22805800-22806721, belonging to the intergenic region (IGR); (4) chr20:21686199-21687689, the gene is PAX1; (5)chr12:101603387-101603933, the gene is SLC5A8; (6) chr11:132952538-132953307, the gene is OPCML; (7)chr19:58609338-58609988, the gene is ZSCAN18; (8) chr12:128751041-128753151, the gene is TMEM132C; (9) chr8:110986113-110986983, the gene is KCNV1; (10)chr9:139780406-139781527, the gene is TRAF2; (11)chr8:105478672-105479340, the gene is DPYS; (12)chr3:5137548-5137977, the region belongs to the intergenic region (IGR), and the reference genome version is hg19.

2. The use according to claim 1, characterized in that The screening for high-grade cervical lesions uses cervical tissue or cervical exfoliated cells as samples.

Images